[QUOTE=dbaugh;276213]If I am using my graphics card for mfaktc how is it that the GPU plays nice by also driving the display? I have heard that CPU onboard graphics is disabled if a video card is detected. Is there a way to trick the computer into using onboard graphics and leave all the GPU for math? Or does the GPU doing double duty not affect TF throughput?(SNIP)
- David[/QUOTE]

I have had various results with the display running mfaktc. At the moment, I've got three instances running, which pretty much maxes out the GPU. There is some lag, compared to running without mfaktc, but it's not really significant typing in a forum. Here in Win7-64 I'd probably shut down at least one of them if I wanted to run Photoshop.

My Asus board has an AMD/ATI onboard chip.....HD4290, I think. When I put in the nVidia card I shut off the onboard video in the BIOS. I had already uninstalled the drivers. I also had to take out the ATI OpenCL driver. It conflicted with the nVidia-based board. You might want to look for video-related settings in your motherboard BIOS to see what those show. You can probably have both display adapters working as long as there are drivers to recognize both of them.

I've thought about going back to the onboard chip to drive the display on my setup, but I haven't tried it yet. It should work.

With mfaktc and Win7-64, on a GTX 460, one of my CPU cores is enough to (barely) saturate it, but I can watch high def movies and play TF2 just fine without lag. It does affect performance, like 20% for the former and 40-50% for the latter, but of course running TF2 isn't trivial.

Thanks Oliver. I didn't mean to imply that you should code a new version. Your contribution to the project has already been very valuable. GPU-TF is the best thing to have happened to the project since the P4 and v21. And you have contributed a lot of crunching too. Right from the days when you gave the old TPR a scare with your march up the TF stats. Old times!

On the AMD 6-core Phenom II, I was able to tell X-windows to run the on-board ATI/AMD video at the same time as the NVidia GPU. But I don't run Photoshop, HD movies, or games on that box at all....and lately, the box itself is sick, so I need to post a "computer diet -- need help!" thread.

Hi Garo,

[QUOTE=garo;276228]Thanks Oliver. I didn't mean to imply that you should code a new version. Your contribution to the project has already been very valuable.[/QUOTE]

I wouldn't be a big deal to write another kernel which takes benefit (can handle) ILP. But I don't think it will be faster (at least on current hardware). Main issue with ILP are the add/sub with carry which are used quiet often in current kernels. Without them I need to do [I]soft carries[/I] and I can't use the full 32 bit per variable for data. This means alot additional shifts, adds and bitmasks. Current kernels do not run totally bad on CC 2.1. So the hypothetical kernels tuned for ILP can better utilize the GPU cores but might be slower anyway.

[QUOTE=garo;276228]And you have contributed a lot of crunching too. Right from the days when you gave the old TPR a scare with your march up the TF stats. Old times![/QUOTE]

How many years ago?
Seems that my password for the arstechnica forum does not work anymore? I wanted to reply on Paulie that I won't bet on Mac support for mfaktc.

Oliver

I think it was at least 7 years ago, maybe more. Doesn't Ars have a lost password link. I'll post that response for Paulie if you like.

Epitaph of the grave of Mr Longbottom..
 

[QUOTE=garo;276322]I think it was at least 7 years ago, maybe more. Doesn't Ars have a lost password link. I'll post that response for Paulie if you like.[/QUOTE]

Ars longa vita brevis

I would like to ask, how to set mfaktc to slow down my computer at least as possible. (my hw: Core 2 Duo E6600@2880 MHz, 4 GB DDR2, GeForce 250 GTS 512 MB, running in 64-bit Linux. I have set it to NumStreams=1, CPUStreams=1, GridSize=1, is that OK?

Use your OS to set the priority of mfaktc as low as possible. And tell it to fix SievePrimes at 10000 (do relatively less sieving out of candidates on the CPU). The other settings you mentioned are also OK.

All this comes at a price in how much work mfaktc does, of course...but the 250 is a relatively low end card these days. You could also see if your card supports CUDALucas and have it do that..that will also slow down the card and therefore the impact on your CPU.

[QUOTE=Wizzard;276533]I would like to ask, how to set mfaktc to slow down my computer at least as possible. (my hw: Core 2 Duo E6600@2880 MHz, 4 GB DDR2, GeForce 250 GTS 512 MB, running in 64-bit Linux. I have set it to NumStreams=1, CPUStreams=1, GridSize=1, is that OK?[/QUOTE]

Something I have been trying in the last couple of days is to set "AllowSleep=1" in mfaktc.ini. It's the last line in the file. It leaves the CPU at a little less than 100%. The difference in mfaktc performance seems to be minimal if you're not doing anything else on the machine. But other apps seem a bit more responsive.

I have never tried setting NumStreams, CPUStreams, or GridSize that low. I have those back at 3,3,3, now. I was running them at 2,3,2 for a while, but it did really slow down mfaktc. I'm not sure it helped that much with other apps.

All this is on a Phenom II 1090T x6, GTX 460, 8GB RAM, with 2 or 3 instances of mfaktc. Right now I'm only running 2 because I'm trying to finish up some partial LL assignments on one core.

EDIT: This is with Win7-64. The priority setting approach I use would not apply in your case. But as Christenson said, CUDALucas has much less CPU load than mfaktc. I bow to his experience regarding the Streams and Gridsize settings.

Don't bow to my experience here..I haven't fooled with numStreams, CPUStreams, etc that much..one core of that 1090T saturates a GTX440 card, which is what I play with....go out and try it...that's what I meant. I forgot about AllowSleep (!) it's a good idea too.

[QUOTE=Christenson;276547]Don't bow to my experience here..I haven't fooled with numStreams, CPUStreams, etc that much..one core of that 1090T saturates a GTX440 card, which is what I play with....go out and try it...that's what I meant. I forgot about AllowSleep (!) it's a good idea too.[/QUOTE]

OK! Go out and try it. That's a good philosophy. You never know what might turn up in a particular setup, with particular personal tolerances.

With the GTX 460, 2 cores of the 1090T keep it pretty busy, though I get a decent increase in productivity with 3 instances/3 cores. As I said, I'm only doing 2 right now so that I have one CPU core to clean up partial LL assignments. The other three are doing P-1.

Thank you very much guys, it helped me a lot. I would also like to try some LL testing on my GPU, but cannot find that precompiled GPULucas for Linux64...

[QUOTE=Dubslow;276217]With mfaktc and Win7-64, on a GTX 460, one of my CPU cores is enough to (barely) saturate it, but I can watch high def movies and play TF2 just fine without lag. It does affect performance, like 20% for the former and 40-50% for the latter, but of course running TF2 isn't trivial.[/QUOTE]

Well I am getting another machine to play around with GPU crunching. It is a 5.27 GHz i7-990X with a GTX 580 video card.[URL="http://www.liquidnitrogenoverclocking.com/trinity_lightning.shtml"][/URL]

I'd like to know if there is any software that COMBINES CPU and GPU computing under one hood. Lots of what I have seen is made for one or the other.

Not yet. It's expected that when the software becomes mature enough and the hardware widespread enough, the various GPU programs will eventually be integrated into Prime95. mfaktc, for instance, is the most developed GPU program yet, but it doesn't even yet have automatic PrimeNet communication. All in due time, I guess.

Also, as for that comp, it seems to me it would be much more cost effective to buy a similar hardware system with a $100-$150 regular liquid cooler, get the same proc to around 4-4.3 GHz, with a 580, for around $2000-$2500, and then get three of those setups, which I'm pretty sure would beat the one above on GIMPS productivity. Unless of course that's not all you're aiming for.

Also keep in mind the Sandy Bridge hex cores are coming out in a couple of weeks, Sandy Bridge has shown anywhere from 10-20% improvement clock per clock over previous Intel architectures. ([URL="http://www.tomshardware.com/reviews/sandy-bridge-core-i7-2600k-core-i5-2500k,2833-2.html"]Here[/URL]: Specifically look at the two audio encoding charts.)

"At the end of the day, we had to scratch our heads and wonder how many folks would be willing to spend almost $700 more on Core i7-990X when Core i7-2600K was already so fast, and priced at $315." -Tom'sHardware

[url]http://www.tomshardware.com/reviews/core-i7-3960x-x79-performance,3026-7.html[/url]
[url]http://www.tomshardware.com/reviews/core-i7-3960x-x79-performance,3026-8.html[/url]
Comparing the new Sandy Bridge hex cores to SB quads and the 990X that LNO is selling.

[QUOTE=SaneMur;276868]Well I am getting another machine to play around with GPU crunching. It is a 5.27 GHz i7-990X with a GTX 580 video card.

[URL]http://www.liquidnitrogenoverclocking.com/trinity_lightning.shtml[/URL]

I'd like to know if there is any software that COMBINES CPU and GPU computing under one hood. Lots of what I have seen is made for one or the other.[/QUOTE]


mfaktc seriously does combine CPU and GPU under one hood -- the CPU feeds factor candidates to the GPU, which then determines whether they are indeed factors. Now, gaming-wise, think of phys-X and SLI. Games that use these are running their physics simulation on one card, and graphics rendering on another, and the CPU ties it all together.

As for cost-effectiveness: Dollar-cost wise, putting together a good power supply and a reasonable motherboard and CPU yourself is probably the fewest dollars per GIMPS CPU/GPU horsepower. Don't fool with overclocking.

[QUOTE=Christenson;276903] Don't fool with overclocking.[/QUOTE]
I find that hard to believe. If you have a decent mobo and the right proc, a decent heatsink isn't too much more money for 20-50% gain in throughput, depending. We being not superstitious, numbers would convince me you are right. Do you have any?

I just find overclocking takes time, and reduces component life significantly, taking up more of my personal time. Thus, cost exceeds benefit. Your mileage may differ.

Well, I'm 18 and only recently into this hardware stuff, and I'd be doing it on Sandy Bridge, which is really easy. I guess my limited experience was just the good stuff :smile:

[QUOTE=Dubslow;276936]Well, I'm 18 and only recently into this hardware stuff, and I'd be doing it on Sandy Bridge, which is really easy. I guess my limited experience was just the good stuff :smile:[/QUOTE]

Sigh — I was 18 once, in 1967....my Intel core-i7 970 was stock at 3.2 GHz and is OC'd to 3.6 with little effort. It has an Asetek closed loop water cooler system and the whole works is in a Raven case of the design where the motherboard is turned 90[SUP]o[/SUP] and the heat exhausts out the top.

[QUOTE=Christenson;276934]I just find overclocking takes time, and reduces component life significantly, taking up more of my personal time. Thus, cost exceeds benefit. Your mileage may differ.[/QUOTE]

You did not count additional power/energy requirement, and also the higher probability of failure (LLERR). Average failure rate as George said somewhere is about 1.8-2% (if I correctly recall, but the absolute value has no importance for the next example). So, in every 100 exponents you return, some of them (say 2, for the given numbers) will have LLERR. When you overclock, you may get the things done few percents faster, but if you return 4-5 LLERR then everything is in vain, not counting additional energy bill.

That was the dark side. The light side is: Of course, no one stops you to try, your mileage may vary, and in fact, if your hardware is pretty good, you can get 10%-30% more work done (and therefore more credit) by overclocking, and without errors.

10% of the price of my medium system is $50-$100. Better to put that together 10 times and buy another system! But have fun with your overclocks.

I got around 33% gains in time per iteration when I added an aftermarket cooler and overclocked my i5-2500k. Went from around .017-.018 iterations a second to .012-.013 for DC testing exponents in the M25000000 range.

What cooler?

Just the Hyper 212. It is really really fantastic for just $25-30. I can post my clock speed when I get home if you want. All I did was tell my bios to overclock me to whatever it thought I should. I run in the mid 60's for the most part.

Oooh, yes, please do. That's what I was looking at for my 2600k. (Also can you tell me your case dimensions, so I can check against mine?)

[QUOTE=KyleAskine;277054]Just the Hyper 212. It is really really fantastic for just $25-30. I can post my clock speed when I get home if you want. All I did was tell my bios to overclock me to whatever it thought I should. I run in the mid 60's for the most part.[/QUOTE]

I certainly endorse the Hyper 212. It didn't get the highest reviews, but you can't beat the price. And the temperature drop was impressive from the stock 1090T cooler.

[QUOTE=Dubslow;277065]Oooh, yes, please do. That's what I was looking at for my 2600k. (Also can you tell me your case dimensions, so I can check against mine?)[/QUOTE]

My case is a little over 6-3/4" from motherboard to side panel. The cooler is about 6-1/2 to the tops of the heat pipes.

Alright. According to CPU-Z I have a bus speed of 100.3 MHz, and a multiplier of 42x, for a core speed of 4214.1 MHz. I am running at an average of around 60 degrees Celsius for each core.

I don't know exact dimensions (I can measure if you want), but I have one in both an old Rosewill minitower case for my Phenom x4, and a newer HAF 932 for my 2500k. It is definitely a bit tall, but if you have a decent width case, it should work. It fits very comfortably in my 932, and incredibly snugly in my old minitower (literally, probably a quarter inch of clearance, if that).

It really does make a world of difference though. I highly recommend it. If you are going to air cool (or probably cool at all), I don't think you can get better performance per dollar.

Hmm. Mine is about 6.5 from the mobo, 6.8 from one side panel to the other. On the other hand, I also keep that side off.

So you get 60C at 4.2GHz 24/7 full load 2500k? Awesome.

[QUOTE=Dubslow;277083]Hmm. Mine is about 6.5 from the mobo, 6.8 from one side panel to the other. On the other hand, I also keep that side off.

So you get 60C at 4.2GHz 24/7 full load 2500k? Awesome.[/QUOTE]

Yup.

I ended up replacing the cooler on my old Phenom x4 because I think I was having heat related issues (not overclocked), as it was running in the 80s if I even ran mprime on two cores, and would crash if I brought up a worker on the third.

After installing the Hyper 212 I can run mprime on all four cores and be in the high 40s or lower 50s. I have since brought it back to three cores on that box, but only because I am running mfakto on the fourth to try and saturate my hd5870 (although mfakto actually takes much closer to two cores).